Would replication still be useful with high write frequency? A different replication strategy would of course be necessary, but communication overhead is a much bigger problem in this case than with cache coherence, so I doubt that the principles used in that case would apply.
jkorn
@rc0303 I do believe that this would be less effective on data that is very frequently written. You'd probably only want to employ this replication strategy on data that you know won't be written to very much, so updates don't have to be broadcasted as frequently.
However, I'm not sure I see your point about communication being a much bigger problem here than with caches - it's a huge problem there too! I think a lot of the same invalidation protocols can be employed here.
Would replication still be useful with high write frequency? A different replication strategy would of course be necessary, but communication overhead is a much bigger problem in this case than with cache coherence, so I doubt that the principles used in that case would apply.
@rc0303 I do believe that this would be less effective on data that is very frequently written. You'd probably only want to employ this replication strategy on data that you know won't be written to very much, so updates don't have to be broadcasted as frequently.
However, I'm not sure I see your point about communication being a much bigger problem here than with caches - it's a huge problem there too! I think a lot of the same invalidation protocols can be employed here.